Transcriptome study exploring the mechanisms linking pollutants and breast cancer aggressiveness

Abstract

Breast cancer (BC) is the most common cancer in women, with metastasis as the leading cause of mortality. Persistent organic pollutants (POPs) may increase BC aggressiveness, but mechanisms remain unclear. This study examined the link between POP exposure and BC progression. We analyzed tumor samples from 89 BC patients in the METAPOP cohort, measuring 42 POPs, including dioxins, polychlorinated biphenyls (PCBs), and polybromodiphenylethers (PBDEs). Transcriptomic analysis identified genes differentially expressed in relation to POP exposure and BC aggressiveness (tumor size, metastatic risk, lymph node involvement, and estrogen receptor status). RNA sequencing of 89 tumors identified 4931 genes differentially expressed with ER status, 283 with tumor size, and 99 with metastasis. Key enriched pathways included immune response, extracellular matrix remodeling, and cell cycle regulation. Gene set enrichment analysis revealed significant overlap in pathways regulating the cell cycle, immune response, and BC hallmarks. Findings suggest that POP exposure alters tumor biology, promoting an aggressive phenotype. This study provides novel insights into the role of environmental pollutants in BC progression, emphasizing the need for further research on how POPs contribute to tumor aggressiveness.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-25528-w.

Introduction

Breast cancer (BC) is the most common malignant tumor in women affecting more than 2 million females in the world each year. BC is currently the second-leading cause of death with more than 670,000 deaths from breast cancer worldwide in 2022, mostly from metastasis (> 90%). Recent findings suggest that environmental pollutants could promote the biological metastatic processes.

Indeed, clinical epidemiological studies have suggested that the mortality caused by BC increases after industrial exposure to 2.3.7.8-TCDD in patients living next to the Chapaevsk or Hambourg factory^1–3. 2.3.7.8-TCDD belong to dioxins, a family of toxic pollutants issued from industrial (combustion, chlorine bleaching, herbicides and pesticides) but also natural processes, such as volcanic eruptions and forest fires. Dioxins are classified as persistent organic pollutants (POPs) because of their impact on human health and of their persistence in the environment and in human/animal tissues. The French METAPOP study was thus designed to evaluate the associations between the concentrations of an extended list of POPs (including dioxins, PCBs and brominated flame retardants) and BC aggressiveness^4,5. Single pollutants and multipollutants models supported a positive association between 2.3.7.8-TCDD and the group of polychlorinated dibenzodioxins with the metastatic cancer risk and tumor size among women with a higher body mass index^4,5.

In vitro studies have shown that these pollutants could contribute to BC metastasis. We identified several cellular and molecular pathways linking POPs treatment and aggressiveness such as (1) maintaining signals for cell proliferation, (2) resisting cell death, (3) promoting neoangiogenesis, (4) EMT and cell invasion, (5) cancer stem cells characteristics, and (6) increasing resistance to treatment⁶. We have shown that 2.3.7.8-TCDD exposure of BC cells cultured in the presence of adipocytes (in a model mimicking the tumor environment) can promote stem cell characteristics (e.g., increased CD24/CD44 ratio) and the appearance of giant cells with cell-in-cell structures which are associated with chemoresistance⁷. Similarly, treatment of this coculture model (BC cells/adipocytes) with cigarette smoke extracts (CSE) increased their stem cell character, and triggered EMT & tumor invasion⁸. CSE could also promote resistance to chemotherapeutic treatment: BC often express estrogen receptors (ER-positive) whose activation favors cell proliferation and can therefore be treated by therapies targeting these receptors. In vitro, it was found that CSE could change BC phenotype from ER-positive cells to ER-negative cells, suggesting that this therapeutical option would be compromised⁸.

The mechanisms linking the exposure to dioxins and BC metastasis remains not fully characterized; nonetheless all POPs associated in the METAPOP study and used in vitro, are ligands of the aryl hydrocarbon receptor (AhR). The activation of the AhR, well-known as a xenobiotic receptor, seems to trigger different key events such as increased inflammation or endothelial migration⁹. Furthermore, the contribution of peritumoral adipocytes to promote a favorable microenvironment to boost the metastasis and progression of tumoral cells, has been suggested due to different signaling pathways supporting the cross-talk^10–12. After contact with BC cells, adipocytes lose their differentiation and acquire a fibroblastic pre-adipocyte phenotype called “cancer-associated adipocytes” (CAA) promoting aggressivity¹³. Moreover, these adipocytes can store pollutants for long period of time¹⁴.

Despite the advent of omic approaches in environmental research to identify potential mechanisms and biological functions, none of previous studies have explored the transcriptome of the tumoral tissue associated to BC metastasis and how POPs influence it¹⁵. In order to gain insights on the impact of POPs on BC cells, we conducted a tumoral tissue transcriptomic study to identify biomarker genes and biological pathways associated with exposure to POPs and BC aggressive characteristics.

Methods

Study population

The present study uses data and biological samples from the METAPOP study described elsewhere^4,5. Briefly, this single-center cohort study included patients diagnosed with breast cancer who were managed at the Department of Gynecological-Oncological Surgery at the Georges-Pompidou European Hospital (HEGP, Paris, France) between December 2013 and November 2017. Eligible participants were women aged 18 years or older with newly diagnosed breast cancer undergoing surgery for uni- or multifocal lesions, with the primary lesion being either palpable or measuring greater than 1 cm in diameter. All participants provided informed consent, and the study received approval from the Comité de Protection des Personnes (the French equivalent of an Institutional Review Board) in 2013. The study is registered on ClinicalTrials.gov under the identifier NCT03788187. All methods were performed in accordance with the relevant guidelines and regulation and in in accordance with the Declaration of Helsinki.

Data collection

Patients underwent either partial or total mastectomy. An adipose tissue sample (1–3 g) was collected approximately 1 cm from the palpable tumor for the quantification of persistent organic pollutants (POPs) and a tumoral sample (1–3 g) was also extracted for transcriptomic analysis. These samples were stored at the hospital’s Biological Resources Center and Tumor Bank Platform (BB-0033-00063). Briefly, data collection included: (i) demographic information, (ii) tumor characteristics, and (iii) tumor extent. Ethnicity was not collected in our work since it does not alter BC prognosis. We did not collect information on environmental exposure since it was objectively measured by the chemical measurement. Gynecologic oncologists collected these data following standardized protocols. Patient information was anonymized using sequential inclusion numbers and recorded in a computerized database.

Details can be found in the supplemental material.

Chemical measurement

The levels of POPs were measured in adipose tissue for each patient. The isolation, detection and quantification of the targeted POPs was measured using ultra-trace methods previously described elsewhere and detailed in the Supplemental material⁵. Forty-two POPs were monitored, among them 17 dioxins (PCDD/F), 16 polychlorobiphenyls (PCB), 8 polybromodiphenylethers (PBDE) and 2,2’,4,4’,5,5’-hexabromobiphenyl (PBB153). The complete list can be found in Supplemental Table S1.

RNA extraction

Tumoral tissue samples were frozen (-80 °C) for every patient. For total RNA extraction, the QIAzol ^® (QIAGEN ^®) lysis reagent was used. The protocol can be found in the supplemental material. The samples were stored at -80 °C. The concentrations and ratios A260/A280nm and A260/A230nm for RNA purity analysis were measured with a Nanodrop One (Ozyme ^®).

Transcriptomic analysis: RNA sequencing

RNA sequencing and analysis was carried out on the GENOMI’C Platform (UDP-8104, Dr Frank Letourneur, Université Paris Cité, Cochin institute, Paris). The RNA samples obtained were analyzed with the Bioanalyzer (Agilent). Only samples with an RIN (RNA Integrity Score) greater than 7 were retained. RNAseq was performed using Illumina sequencing technologies. Sequence alignment was performed using the reference genome from the “Ensembl” database with the STAR program.

Data analysis

Transcriptomic data was filtered to keep at least 10% of counts. Count data was then normalized and scaled before the statistical analysis. Descriptive statistics of sociodemographic, clinical and exposure data was conducted summarizing and comparing the distributions of continuous data using Mann-Whitney and Student t-tests, whereas categorical data was compared between groups using Fisher tests. Principal component analysis (PCA) was used to explore the variation due to potential covariates. Hierarchical clustering on principal components identified in the PCA on transcriptomic data, was conducted to identify the main clusters of observations and to explore the sources of variation. To this end, data related to clinical, sociodemographic and exposure variables, were compared using univariate statistics between cluster groups. We then developed a “meet-in-the-middle” statistical workflow to identify differentially-expressed genes (1) related to BC aggressiveness (number of involved nodes, tumor size, presence of metastasis); (2) levels of POPs and (3) overlapping between POP exposure and aggressiveness. Cancer sub-types and aggressiveness hallmarks included tumor size, metastatic status, number of affected lymph nodes and expression of estrogen receptor alpha. Considering the large number of POPs and high correlation between congeners within families, we conducted a clustering based on latent variables to support the dimension reduction into 4 main k-groups, depicted by the sum of PCDDs, PCDFs, PCBs and PBDEs congeners¹⁶. Differential Gene Expression (DGE) analysis based on a negative-binomial Wald test with normalized counts, was conducted using the DESeq2 pipeline and implemented with the “DESeq” function within the DESeq2 (v.1.44.0) R package¹⁷. As illustrated by the method developers, we considered a q-value < 0.1 (i.e., p-values adjusted for multiple testing using the Benjamini-Hochberg procedure) to maximize sensitivity in exploratory analyses with moderate or small sample sizes¹⁷. To gain insight into the biological mechanisms, gene set enrichment analyses (GSEA) were performed with the output of DGE analysis and integrated with the Kyoto Encyclopedia of Genes and Genomes (KEGG) and Gene Ontology biological function databases implemented with the gage R Bioconductor package and ShinyGO v0.75^18–21.

We first conducted an exploratory PCA on different characteristics such as the outcomes (ER positivity, metastasis, tumor size, number of affected nodes), chemical exposure levels and patient characteristics (age, BMI, smoking, menopause or family history of BC). The goal of an exploratory PCA is to reduce the dimensionality of a dataset while preserving its most important variability. It transforms correlated variables into uncorrelated principal components, helping to simplify data, identify patterns, visualize trends, and detect outliers.

Results

Description of study population

Transcriptomic profiles were available for 89 women out of the ninety-one included in the METAPOP study with measured POP levels. The mean age was 62 years old (+/- 14, range 37–93 years old), 73% were menopaused (65/89) and the mean BMI was 24.4 kg/m² (range: 15.7–57). The mean tumor size was 24.8 mm (+/- 13.6 mm, range: 4–75 mm), the majority of patients had tumors measuring > 2 cm (52.8%) and most had an invasive ductal carcinoma (77.5%). The most frequent immunohistochemistry finding was luminal B (positive hormonal receptors, low proliferation, Ki67) (51/89), followed by luminal A (positive hormonal receptors, high proliferation, Ki67) (24/89), triple negative (negative hormonal receptors and Her2 not over expressed) (11/89) and Her2+ (Her2 over expressed) (3/89) cancers. Thirty-six patients (40.4%) presented a metastatic cancer and 75 expressed estrogen receptors (ER-positive). Demographic characteristics are presented in Table 1 and patients with and without metastasis and with and without hormonal receptors are compared in Tables S2 and S3. No demographic data differed significantly for these outcomes and therefore the entire population was included.

Table 1.

Summary distributions and frequencies of sociodemographic and clinical variables from women included in the present study (N = 89).

	Total N = 89 (%)
Age (years) mean (SD)	63.2 (14.3)
Body mass index (kg/m2) mean (SD)	24.4 (4.06)
Smokers (%)	16 (18.0%)
Family history of breast cancer (%)	20 (22.5%)
Menopause (%)	65 (73.0%)
Parity
0	21 (23.6%)
1	21 (23.6%)
2	25 (28.1%)
≥ 3	22 (24.7%)
Breastfeeding	41 (46.1%)
Estrogen receptor +	75 (84.3%)
Metastatic cancer	36 (40.4%)
Lymph nodes affected
0	53 (59.6%)
1	24 (27.0%)
≥ 2	12 (13.5%)
Tumor Size
< 20 mm	41 (46.1%)
> 20 mm	48 (53.9%)

Exploratory analysis

Exploratory PCA of count data showed that the two first components accumulated 17% of total variance. Covariates such as age, BMI, smoking, menopause or family history of breast cancer showed little discrimination of individuals on these first components (Figure S1 and S2). Hierarchical clustering on the principal components of count data showed two main clusters of 50 and 39 women (Figure S3), with statistically significant differences on estrogen receptor alpha status (2 vs. 12, Fisher test, p = 0.0008) and marginally different levels of polybrominated biphenyl 153 (PBB153), a flame retardant (mean 0.1 vs. 0.6 ng/g, t-test, p = 0.068) (Table S4). We therefore decided to study PBB153 in the following analysis. TCDD and PBB153 exposure levels showed little discrimination (Figure S4).

In order to reveal potential associations between genomic variables and possible cofounding variables, we conducted DGE analysis on the covariates. Age and BMI were the variables that seemed to influence the majority of genes (104 and 91 genes respectively at a q-val < 0.1, Figure S5). On this basis, we included age and BMI in the subsequent DGE analysis of BC hallmarks and POP levels as covariates.

Breast cancer-related transcriptomic profiles

All the breast cancer variables showed differential expressions (Fig. 1). For instance, 69 genes were significantly upregulated among cases with metastatic risk whereas 30 genes were downregulated (Fig. 1A). The number of affected lymph nodes revealed 104 differentially regulated genes in cases with 2 or more affected lymph nodes were compared to those without any node affected (Fig. 1C). By far, the largest difference in the number of genes affected was related to the presence or absence of the estrogen receptor expression in cancer cells. Indeed, 4931 genes were differentially expressed in the tumoral tissue depending on the estrogen receptor status (Fig. 1D). Tumors with the largest size (> 2 cm) exhibited 283 genes downregulated compared with the smaller ones (Fig. 1B). The lists of top ranked genes can be found in the supplemental Tables S5-S8.

Fig. 1.

Volcano plot for differential gene expression in patients in the METAPOP cohort. In gray, non-significant gene expression, in green significant change in gene expression on the fold change, in blue significant change in gene expression on the q-value, in red significant change in gene expression on the fold change and p-value, using a log₂ Fold change cut off of 0.5 and q-value cut off of 0.1. Tumor size (> 2 cm) and estrogen receptor status have the largest number of genes differentially expressed.

Pathway analysis using GSEA showed that 6 GO biological processes and 2 GO molecular functions involved in extracellular matrix organization and cytokine activity were differentially expressed according to the presence of breast cancer metastasis (Table S9). Tumor size was associated to the upregulation of 200 GO biological processes and 4 KEGG metabolic pathways related to cell cycle and function (Figure S6). Similarly, the number of affected lymph nodes was associated with an upregulation of cell cycle and RNA splicing GO and KEGG pathways among others (Figures S7). In turn, the presence of estrogen receptor in cancer cells was associated with 11 downregulated KEGG metabolic pathways and 297 biological processes related to immune function, hypoxia, spliceosome and cell function (Figures S7).

POP-related transcriptomic profiles

Due to our previous findings and available literature, we chose to focus on 2.3.7.8-TCDD to gain insight on potential mechanisms supporting the links with adverse cancer-related outcomes^5,7. We conducted a DGE associated with PBB153 levels based on the pilot exploratory clustering analysis (See details in Sect. Exploratory analysis). In addition, we conducted a DGE with the 4 main POP groups depicted by the dimension reduction CLV method, globally grouping the congeners in PCDDs, PCDFs, PCBs and BFRs. The results are summarized in volcano plots in the Fig. 2. Using a threshold of significance of q < 0.1, we found 40 genes differentially associated to 2.3.7.8-TCDD levels and 126 genes associated with PBB153 levels (Fig. 2A and B). The DGE on chemical clusters showed major contribution of PCBs on the transcriptome (276 statistically associated genes compared to 36–81 genes for the other chemicals (Fig. 2C and F).

Fig. 2.

Volcano plot for differential gene expression in patients in the METAPOP cohort. In gray, non-significant gene expression, in green significant change in gene expression on the fold change, in blue significant change in gene expression on the p-value, in red significant change in gene expression on the fold change and q-value, using a log₂ Fold change cut-off of 0.5 and p-value cut off of 0.1.

Despite the large number of genes associated with PCBs, only one upregulated KEGG metabolic pathway was significant in the GSEA analysis (hsa03040 Spliceosome). In turn, the cluster of BFRs exhibited a relevant number of downregulated GO biological functions (n = 58) and 4 KEGG metabolic pathways such as protein processing (hsa04141), oxidative phosphorylation (hsa00190) or cell cycle (hsa04110)²². The list of genes and pathways can be found in the Supplementary material (Table S10-15 and Figure S9-10).

Genes intersecting exposure to pops and breast cancer outcomes

The summary of number of genes differentially regulated by BC characteristics and exposure to POPs can be found in Table 2. Several genes were commonly associated across chemicals and BC characteristics. For instance, the expression of the fat storage inducing transmembrane protein 2 (FITM2) is negatively associated with TCDD (Log2FC=-0.80) and tumor size (Log2FC=-1.25) and positively with PBB153 (Log2FC = 0.92), PCBs (Log2FC = 0.92) and metastatic risk (Log2FC = 1.04). Glial cell derived neurotrophic factor (GDNF) was positively associated with metastatic cancer (Log2FC = 3.85) and negatively with tumor size (Log2FC=-4.09), TCDD (Log2FC=-2.40), PBB153 (Log2FC=-1.72), PCDFs (Log2FC=-2.70) and PCBs (Log2FC=-2.26). UDP glucuronosyltransferase family 2 member B4 (UGT2B4) appeared to have a lower expression when exposed to higher levels of TCDD (Log2FC=-5.03), PBB153 (Log2FC=-3.03), PCDFs (Log2FC=-4.27) and PCBs (Log2FC=-3.81) and the number of lymph nodes (Log2FC=-18.65)²³. In turn, cytokine like 1 (CYTL1) had a higher expression among metastatic cancer (Log2FC=-1.91) and a lower expression when TCDD levels were increased (Log2FC = 1.09) and PCBs (Log2FC = 1.20). The gene encoding for aquaporin 4 (AQP4) had a higher expression in case of higher levels of PCDDs (Log2FC = 4.03), PCBs (Log2FC = 3.41) and the number of affected lymph nodes (Log2FC = 5.05) and metastatic risk (3.48) but had a lower expression in case of ER + cancer (Log2FC=-4.83)²⁴. The complete list can be found Tables S16-S19 and details are shown Fig. 3.

Table 2.

Number of genes overlapping exposure to pops and breast cancer hallmarks. The detailed list of genes can be found in the supplemental tables S16-S19.

	TCDD	PBB153	PCDDs	PCDFs	PCBs	BFRs
Metastatic risk	6	18	6	6	26	16
Oestrogen receptor +	15	55	14	29	66	23
Lymph nodes	4	5	4	3	19	13
Tumor size	8	26	4	7	31	14

Fig. 3.

Venn diagrams showing the number of differentially regulated genes among breast cancer outcomes (A metastatic risk, B ER-positivity, C lymph nodes and D tumor size) and also showing differential expression by exposure levels of pollutants, with specific overlap among them. A significance level of q-val < 0.1 was considered to select the genes using the Deseq2 analysis.

Discussion

In this work, we assessed the crosstalk between internal exposure levels of POPs, the patterns of gene expression in the tumoral tissue and BC outcomes through a transcriptome-wide study. We aimed to identify genes and biological pathways associated with exposure to POPs and aggressive BC subtypes. While several studies have focused on the tumoral transcriptome, none have studied the cross talk between pollutants, transcriptomic modifications in the tumor and BC aggressiveness^25,26.

Our transcriptomic analysis showed a differential gene expression in tumoral tissue for BC outcomes such as: metastasis (99 genes differentially regulated), ER-positivity (4931 genes), tumor size (> 2 cm) (283 genes) and number of involved lymph nodes (2 or more) (104 genes). Pathway analysis using KEGG and GO showed: i) for metastasis, pathways mainly involved in the remodeling of the extracellular matrix and in cytokine activity, ii) for tumor size spliceosome, cell cycle and function, iii) for number of nodes cell cycle, spliceosome and cell function and iv) for ER-positivity immune function, hypoxia, spliceosome and cell function. Modification of the spliceosome in the tumoral tissue, seems to play a key role in promoting BC aggressiveness; indeed, aberrant splicing events have been implicated in cancer progression by altering the expression of genes involved in cell cycle control, apoptosis, and metastasis^27–29. There is little data on the impact of higher expression of spliceosomal RNA yet, it has been shown that alternative splicing events occurring in the tumor microenvironment, could promote cancer progression³⁰. U1 and U2 spliceosomal RNA were overexpressed in the tumoral tissue and associated with poor BC outcomes. U1 snRNA recognizes the 5’ splice site, initiating the splicing process, while U2 snRNA is crucial for spliceosome assembly and catalytic activity. In BC, alternative splicing in the tumor could also modify the tumor microenvironment immune response³¹. The tumor also seems to confer an aggressive phenotype through the release of inflammatory mediators: indeed, several cytokines (CYTL1), chemokines (CCL23, CX3CL1), interleukins (IL1RAPL1, IL 20 and IL27RA), tumor necrosis factors (TNFRSF11A), metalloproteinase (MMP14) and bone morphogenic proteins (BMPR1B, BMPER) were increased according to the tumor outcome. Inflammation can in turn activate epithelial to mesenchymal transition (EMT) in BC cells and degrade the extracellular membrane to favor invasion and metastasis^32–35.

To the best of our knowledge, this is the first time that modifications of the BC tumoral transcriptome associated with exposure to persistent pollutants (POP) have been studied in a cohort of patients through a complete RNA-sequencing. Forty-two organic chemicals were assayed in patients from the METAPOP cohort. Among them 17 dioxins (PCDD/F), 16 polychlorobiphenyls (PCB), 8 polybromodiphenylethers (PBDE) and 2,2’,4,4’,5,5’-hexabromobiphenyl (PBB153) were found in the AT surrounding the tumor. We chose here to dose pollutants in the peritumoral adipose tissue due to its capacity to store chemical products. Indeed, adipocytes could play a role in the storage of environmental pollutants^14,36: since the majority of pollutants are lipophilic, they can be stored in the adipose tissue and released chronically, at low levels over long periods of time^37,38. The first cells to be affected are those closest to adipocytes, including breast cancer cells within a tumor. Our transcriptomic analysis showed a differential gene expression for 2.3.7.8-TCDD (40 genes differently expressed), PBB153 (126 genes), PCBs (276 genes), PCDD (44 genes), PCDF (70 genes), PBDE (35 genes). Pathway analysis using KEGG and GO showed: i) for PCDD, pathways mainly involved in cell cycles, for PCBs spliceosome, for BFRs oxidative phosphorylation and cell cycle. We were not surprised to find modifications in the cell cycle since the aryl hydrocarbon receptor binds dioxins and alters cell cycle³⁹.

Finally, we analyzed genes differentially expressed by the tumor at the intersection between BC outcomes and exposure to POPs. Fat storage inducing transmembrane protein 2 (FITM2) seems to be a recurring gene modified, according to the cancer outcome and the POPs. FITM2 is primarily localized in the endoplasmic reticulum and plays a crucial role in lipid droplet formation within cells⁴⁰. We show that the expression of FITM2 is negatively associated with TCDD, the most toxic dioxin, whose levels are associated with BC aggressiveness. The decreased expression of FITM2 could lead to reduced lipid droplet formation, which might impair the ability of the tumor to store lipids effectively, leading to an altered metabolic function, potentially influencing tumor progression. Indeed, the impaired ability to store lipids could trigger enhanced lipolysis and the release of free fatty acids (FFAs) into the tumor microenvironment^40,41. The main microenvironment of BC cells is adipocytes. On the long term, a decreased lipid storage capacity may increase the availability of FFAs for β-oxidation, promoting a shift from glycolytic metabolism (Warburg effect) to lipid oxidation⁴². In summary, these FFAs can serve as an energy source for cancer cells, promoting tumor growth, survival, and metastatic potential by enhancing mitochondrial oxidative metabolism in tumor cells. In summary, the altered expression of FITM2 might indirectly contribute to BC progression by providing a readily available energy source for these rapidly dividing cells. The tumor microenvironment when exposed to these pollutants could alter BC lipid metabolism⁴³.

Likewise, glial cell derived neurotrophic factor (GDNF) is primarily known for its role in neuronal survival and development, but it has also been implicated in cancer biology, including breast cancer^44,45. Studies suggest GDNF can activate pathways that make breast cancer cells, particularly estrogen receptor-positive (ER+) ones, resistant to hormone therapy like tamoxifen⁴⁶. Furthermore, adipose-derived stem cells present in the micro environment could secrete GDNF as a potent angiogenic factor independent of vascular endothelial growth factor (VEGF), thus promoting endothelial network formation in liver cancer⁴⁷.

Also, UDP-glucuronosyltransferases (UGTs) are a family of enzymes involved in the detoxification and metabolism of various endogenous and exogenous compounds by glucuronidation. It has been studied in chemoresistance and ER-metabolism^48,49. UGTs mediate the inactivation and excretion of lipophilic drugs by facilitating the conjugation of glucuronic acid, thereby increasing water solubility and promoting elimination^48,50,51. Moreover, UGT1A1, UGT1A3, and UGT2B15 are involved in the glucuronidation of estradiol and estrone, converting these active estrogens into their inactive glucuronide conjugates. Genetic polymorphism in these UGTs can affect estrogen levels in breast tissue, influencing cancer progression⁴⁹. However, there is scarce data on UGT2B4, studied here.

Finally, some studies suggest 4 aquaporin (AQP4), increased here, might be involved in BC cell migration and invasion by releasing water from the cell, modifying the cytoskeleton and decreasing cell adhesion^52–54. More interestingly, in colon cancer, AQP9 in peritumoral macrophages, transports lactate secreted by the cancer cell, which in turns transforms the macrophages into polarized tumor-associated macrophages⁵⁵. Furthermore, AQP3, is upregulated by the aryl hydrocarbon receptor (AhR) upon exposure to environmental pollutants like TCDD. This upregulation enhances cell migration, and could explain cancer progression⁵⁶.

The main limitation of our work lies in the relatively small sample size, which limits the statistical power and the ability to fully use covariates to inspect interactions or stratifications^57,58. As such, this work should be interpreted as a hypothesis-generating exploratory analysis, rather than a confirmatory one. To maximize sensitivity in this context, we adopted a q-value threshold of < 0.1 for differential expression, as recommended for exploratory transcriptome-wide analyses in small cohorts. Nonetheless, we acknowledge that future studies with larger sample sizes will be essential to validate these findings using more stringent statistical thresholds (e.g., q < 0.05), improve adjustment for covariates, and allow for more robust integrative analyses. In this study, we focused on identifying the overlap between genes differentially regulated by pollutant exposures and breast cancer outcomes, and used pathway-level enrichment (GSEA) to gain biological insight. While additional analyses such as co-expression network reconstruction (e.g., WGCNA or Cytoscape-based hub gene identification) could offer further mechanistic understanding, we considered that our sample size was insufficient to ensure the reliability and reproducibility of such network models. These approaches may be more appropriate in future studies with greater statistical power or in meta-analytic designs. Finally, we carried out an association study and reverse causality is also possible, for example, pollutants altering cancer cell properties which in turn alter adipose tissue pathways. The latter hypothesis seems unlikely based on our previous experimental studies. Our results could not be validated on a Western blot analysis due to the size of the tissue samples.

In conclusion, this is the first work aiming to gain insight in the potential crosstalk between POP, tumor tissue modifications and BC. The identification of biomarkers (deregulated or not by POPs) in tumor cells could represent a therapeutic avenue in the near future to sensitize tumor cells to the effect of treatments that affect them directly. Our study is in line with this approach, while considering the environmental component, bearing in mind that a large number of risk factors for breast cancer and its progression are still unknown.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Louise Benoit : Conceptualization, Data Curation, Formal Analysis, Investigation, Methodology, Software, Validation, Visualization, Writing – Original Draft Preparation, Writing – Review & EditingGerman Cano-Sancho : Conceptualization, Data Curation, Formal Analysis, Methodology, Software, Visualization, Writing – Original Draft Preparation, Writing – Review & EditingCéline Tomkiewicz : Conceptualization, Data Curation, Formal Analysis, Funding Acquisition, Supervision, Writing – Review & EditingAnne-Sophie Bats : Conceptualization, Investigation, Methodology, Supervision, Validation, Writing – Review & EditingNathalie Douay-Hauser : Conceptualization, Data Curation, Formal Analysis, Writing – Review & EditingBruno Le Bizec : Conceptualization, Data Curation, Formal Analysis, Writing – Review & EditingMaxime Delit : Conceptualization, Data Curation, Formal Analysis, Writing – Review & EditingJean-Philippe Antignac : Conceptualization, Investigation, Methodology, Supervision, Validation, Writing – Review & EditingRobert Barouki: Conceptualization, Funding Acquisition, Investigation, Methodology, Project Administration, Supervision, Validation, Visualization, Writing – Review & EditingXavier Coumoul: Conceptualization, Funding Acquisition, Investigation, Methodology, Project Administration, Supervision, Validation, Visualization, Writing – Review & EditingMeriem Koual: Conceptualization, Data Curation, Funding Acquisition, Investigation, Methodology, Project Administration, Resources, Supervision, Validation, Visualization, Writing – Review & Editing.

Funding

This work was supported by PNRPE (METAPOP project, n° 11-MRES-PNRPE-5-CVS-031 Chorus n°2100532530); Institut National de la Santé et de la Recherche Médicale (INSERM); Université Paris Cité (UPCité); Assistance Publique des Hôpitaux de Paris (APHP).

Data availability

The datasets generated and/or analysed during the current study are available in the arrayexpress repository, fgsubs #820429.

Declarations

Competing interests

The authors declare no competing interests.

Ethics approval and consent to participate

All patients gave informed consent to participate in the study. The study was approved by the ‘Comité de Protection des Personnes’ in 2013 [French equivalent of an Institutional Review Board (IRB)] and is registered in clinicaltrial.gov (ID NCT03788187).